機車偵測、追蹤及對應

Scooter Detection, Tracking and Ego-Vehicle Reaction for Self-driving Vehicles

胡家睿 ； 李則霖 ； 詹巧同

《機械工業雜誌》 445期 (2020/04) Pp. 35-42

摘要

在亞洲區發展自主駕駛的難度較歐美地區高，因為亞洲地區的交通發展往往因為地狹人稠而發展出高機動性的機車文化使得交通較複雜。因此為了在台灣發展自主駕駛技術，如何對應機車變成一重要的課題，工研院機械所去年已申請到國內第一張開放場域測試車牌，已於新竹南寮漁港週邊進行複雜交通測試，其中就包含多樣的機車行為應對情境，因此本文將就自主駕駛車如何針對機車之相關經驗來進行介紹。

關鍵字

自主駕駛車 ； 偵測 ； 追蹤

並列摘要

Enabling self-driving vehicles in Taiwan would be more challenging than other areas due to its high complexity of crowded scooter traffic environments. As a result, to deal and react with these scooters on roads becomes an important capability measure for autonomous driving systems. The self-driving vehicles group at MMSL, ITRI obtained the first self-driving car license in the NanLiao open field last year and had collected a number of cases containing special behavior of scooters. This article will reveal some of our approaches for dealing with scooters on public roads.

並列關鍵字

Autonomous vehicle ； Detection ； Tracking

參考文獻

Redmon, Joseph,Farhadi, Ali(2018).YOLOv3: An Incremental Improvement.,未出版.

Qi, Charles R.,Su, Hao,Mo, Kaichun,Guibas, Leonidas J.(2017).PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation.CVPR 2017.(CVPR 2017).

CS468: 3D Deep Learningon Point Cloud Data, Hao Su, Stanford University,2017, http://graphics.stanford.edu/courses/cs468-17-spring/LectureSlides/L16%20-%203d%20deep%20learning%20on%20point%20cloud%20(analysis)%20and%20joint%20embedding.pdf

Tang, Yichuan Charlie,Salakhutdinov, Ruslan(2019).Multiple Futures Prediction.Conference on Neural Information Processing Systems.(Conference on Neural Information Processing Systems).

Argoverse Motion Forecasting Competition, https://evalai.cloudcv.org/web/challenges/challenge-page/454/evaluation